This repository contains utilities for testing generalized relative pose estimation in the context of multi-camera SLAM initialization. The goal of the project is to bring real-world and synthetic datasets together with relative pose solvers to provide a benchmarking framework for the latter.
sudo apt install git g++ cmake libgflags-dev libgoogle-glog-dev libfmt-devThe OpenGV solver initialization pipeline requires a version of OpenCV from 4.4. Please install it from here and provide the installation directory to CMake by setting the additional flag -DOPENCV_INSTALL_DIR=<your OpenCV install dir>
This project also requires Ceres Solver. The installation instructions can be found here.
mkdir build
cd build
cmake .. -DCMAKE_BUILD_TYPE=RelWithDebInfo
cmake --build . -- -j6To build a specific demo you can run
cmake --build . --target demo -- -j6where demo stands for the required binary. Both commands will also build OpenGV, which may take substantial time and may even freeze the machine if built in Debug mode.
This project is licensed under the terms of the MIT license. For more information, please check the LICENSE.txt file.